Instructions to use lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ")
model = AutoModelForMultimodalLM.from_pretrained("lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ

SGLang

How to use lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ with Docker Model Runner:
```
docker model run hf.co/lightblue/Karasu-Mixtral-8x22B-v0.1-AWQ
```

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

4bit AWQ version of the lightblue/Karasu-Mixtral-8x22B-v0.1 model.

Quantized using the following code:

from awq import AutoAWQForCausalLM
import pandas as pd
from transformers import AutoTokenizer
from tqdm.auto import tqdm

pretrained_model_dir = '/workspace/llm_training/axolotl/mixtral_8x22B_training/merged_model_multiling'
quantized_model_dir = '/workspace/llm_training/axolotl/mixtral_8x22B_training/merged_model_multiling-awq'

# The samne dataset as in lightblue/gpt4_conversations_multilingual
df = pd.read_json(
    "/workspace/llm_training/axolotl/mixtral_8x22B_training/sharegpt4_multilingual.json", 
    lines=True)

role_map = {
    "human": "user",
    "gpt": "assistant",
}

df["messages"] = df.conversations.apply(lambda x: [{"role": role_map[y["from"]], "content": y["value"]} for y in x])

tokenizer = AutoTokenizer.from_pretrained(pretrained_model_dir, use_fast=True)
examples = [
    tokenizer.apply_chat_template(
        x, tokenize=False, add_generation_prompt=False
    ) for x in tqdm(df["messages"])
]

model_path = '/workspace/llm_training/axolotl/mixtral_8x22B_training/merged_model_multiling'
quant_path = '/workspace/llm_training/axolotl/mixtral_8x22B_training/merged_model_multiling-awq'
quant_config = { "zero_point": True, "q_group_size": 128, "w_bit": 4, "version": "GEMM" }

# Load model
model = AutoAWQForCausalLM.from_pretrained(model_path)
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)

# Quantize
model.quantize(tokenizer, quant_config=quant_config, calib_data=examples)

# Save quantized model
model.save_quantized(quant_path)
tokenizer.save_pretrained(quant_path)

Downloads last month: 3

Safetensors

Model size

141B params

Tensor type

I32

F16